Saline Water ConversionâII

10 Operating Experience on a Large Scale Electrodialysis Water-Demineralization Plant O.

B.

VOLCKMAN

1

Process Development Division, National Chemical Research Laboratory, Council for Scientific and Industrial Research, Pretoria, South Africa A n electrodialytic w a t e r

demineralization

plant,

w i t h multiple units using the " s h e e t - f l o w "

prin-

Downloaded by UNIV OF BATH on June 27, 2016 | http://pubs.acs.org Publication Date: January 1, 1963 | doi: 10.1021/ba-1963-0038.ch010

ciple, w a s d e s i g n e d to produce 2,880,000 gallons per d a y of w a t e r with 525 p.p.m. of total

dis-

s o l v e d solids f r o m a 3 1 0 0 - p . p . m . TDS mine w a t e r in South A f r i c a .

The plant o p e r a t e d f o r o v e r 18

months, but at not g r e a t e r than

71%

of

rated

capacity, a n d then o n l y w i t h a 2 8 4 3 - p . p . m .

TDS

feed

low

water

and

71.4%

TDS

removal.

The

capacity w a s m a i n l y d u e to excessive p o l a r i z a t i o n and intercompartmental leakage. on full scale units s h o w e d could be e l i m i n a t e d .

Extensive trials

that these difficulties

Desalting costs r a n g e d f r o m

an actual $ 0 . 4 2 6 (U. S.) per 1000 gallons at 7 1 . 4 % TDS r e m o v a l

and

1,540,000-gallon-per-day

out-

put to a n estimated $ 0 . 2 9 7 (U. S.) p e r 1000 g a l l o n s at 8 2 . 1 % TDS r e m o v a l a n d 2,880,000-gallon-per-day

output.

I l l ork on saline water conversion using the electrodialysis process i n South A f r i c a was outlined b y the author i n 1957 ( 7 ) , including plans for a 2,880,000-gallonper-day ( U . S. gallons) desalting plant at W e l k o m , Orange Free State, to desalt a highly saline mine water of approximately 3100 p . p . m . total dissolved solids (3000 p.p.m. N a C l ) to 535 p . p . m . total dissolved solids (500 p . p . m . N a C l ) ( F i g ures 1, 2, and 3 ) . Experience gained from operation of that plant is reported here. T h e reasons for erecting the plant, and the design basis, have been described i n detail (1, 2,4, 5,9). T h e decision to construct the large plant was taken b y the mining companies w i t h f u l l knowledge that the urgency of the water-disposal situation at that time w o u l d permit only very limited prototype tests. This meant that the large plant w o u l d probably have to be operated, for the first 2 years at least, largely as a development plant, since many factors i n the design and operation of large electrodialysis plants h a d not been fully investigated. Some of the operating expenses w o u l d be met b y the sale of the desalted water produced, but it was thought unlikely that the venture w o u l d make a profit i n the earlier stages. Present address, D e p a r t m e n t of C h e m i c a l E n g i n e e r i n g , U n i v e r s i t y of the W i t w a t e r s r a n d , Johannesburg, South A f r i c a 1

133

Saline Water Conversion—II Advances in Chemistry; American Chemical Society: Washington, DC, 1963.


134

A D V A N C E S IN

Figure 1.

CHEMISTRY SERIES

Free State Geduld electrodialysis plant

Dam for 1,000,000 gallons of raw water Right. Filter plant building Rear. Electrodialysis plant building

The plant was principally an insurance against loss of mining production from an inability to dispose of large volumes of saline water. As events developed, severe engineering difficulties were experienced, while the magnitude of the water-disposal problem progressively decreased. Whereas i n 1956, when the project was first considered, the salinity of the mine water from some mines was as high as 4000 p . p . m . T D S , the average value of the water to be treated by the Free State G e d u l d plant had fallen to 3100 p.p.m. T D S i n 1958 and to 2700 b y early 1961. U p to October 1960 the output from the plant of desalted water per unit h a d not exceeded 71.4% of design and this output was only achieved w i t h a product water of 813 p . p . m . T D S . T h e reasons for low production per unit were high electrical resistance due to polarization, sludge formation on the membranes (not scaling), and leakage from the diluting to the concentrating streams. T h e polarization difficulties were eventually overcome ( 1 0 ) . Considerable work was done in the plant and i n the laboratories at Pretoria to overcome the leakage problem (11). It was shown that this problem could be overcome, but because of the additional capital investment required and the fact that the water-disposal problem was rapidly becoming less serious, it was not possible to confirm the findings by longscale tests on the plant. In February 1961, the mining companies decided to suspend operations of the Free State G e d u l d M i n e plant pending further work by the South African Council for Scientific and Industrial Research aimed at reducing costs. A t that stage the plant could have been relied upon to deliver 1,500,000 gallons per day of Saline Water Conversion—II Advances in Chemistry; American Chemical Society: Washington, DC, 1963.


J O . VOLCKMAN

Electrodialysis

Water-Demineralization

Plant

135

525 p.p.m. T D S water, but at a somewhat higher cost than 30 cents per 1000 gallons. As the mines buy their water at R 0.25 ($0.30) per 1000 gallons, they were not prepared to continue operating a plant producing water at a higher cost. A t the time of writing (November 1962) the future of the plant is under final review. A s the amount of water from the mines has decreased to the point where the water can be evaporated from the existing surface evaporation dams, it is probable that the plant w i l l be dismantled. If, i n 1956, the geologists could have predicted the present water situation, the plant w o u l d probably not have been built. Some of the difficulties experienced were expected. They were inherent i n the decision to use large units after only limited prototype tests. These difficulties were largely mechanical and were particularly associated w i t h the design, materials of construction, and fabrication of press components. Other difficulties, however, could have been found only b y prolonged operation i n the field; typical were difficulties relating to adequate removal of the suspended clay i n the raw water, sludging of the membrane compartments with the clay, and failure after prolonged use of certain press components and piping. Nevertheless, the plant has shown without doubt that the electrodialysis process can, under correct operating conditions, be very reliable; membrane packs of large dimensions and with large numbers of membranes (up to 200) can easily be assembled and handled; high voltages (450 to ground) can be used without mishap to operating staff, or causing excessive current leakage or corrosion of surrounding structures or equipment; and simple automatic control suffices where continuous, as against batch, desalting is used. Some Basic Design Considerations The Free State G e d u l d plant was based primarily on what is sometimes described as the sheet-flow, intermembrane spacer, or full-flow type of apparatus,

Figure 2.

Section of electrodialysis presses



136

ADVANCES

Figure 3.

IN CHEMISTRY SERIES

Top of electrodialysis units

as opposed to the tortuous path type. Considerations leading to the design of the plant are given by the South African Council for Scientific and Industrial Research (SACSIR) water desalting team (9); progress is reviewed by Wilson (8) and Volckman (6). As a guide to sizes of existing sheet flow types of electrodialysis plants, Table I compares the membrane area of the F S G plant with those of the next largest commercial plants known to be in commercial operation in early 1961, and which were operating with multiple packs in a press. In considering the design of the plant two related factors had to be considered very carefully. The advantage, if any, of using a few large electrodialysis units rather than a multiplicity of smaller units (the term "unit" is here taken as a membrane press, whether consisting of one or a number of membrane packs). The plant uses the principle of multiple packs in each electrodialysis unit or "press." The type of membranes that should be used, and the practical operating lifetime that could be expected from them (very few commercial membranes, especially of large size, had been extensively field-tested in 1956). One of the severely limiting prerequisites of the plant was that the cost of desalting the water from 3000 p.p.m. NaCl (approximately 3100 p.p.m. T D S )


10. VOLCKMAN T a b l e 1.

Electrodialysis


FSG


137

C o m p a r i s o n of Free State G e d u l d Plant with O t h e r I n t e r m e m b r a n e Spacer Plants

Over-all membrane dimensions, inches Over-all membrane area, sq. ft. % effective membrane area No. of electrodialysis presses on line (excluding spares) No. of packs per press No. of membranes per pack Total membrane area of plant (excluding spare units), sq. ft. Effective membrane area of plant (excluding spare units), sq. ft. A.

Plant

85y

4

A

x 25y 15.33 67

4

59'A X 15»A 6.50 61

B 397

2

X 4 59

8 10 200

1 1 200

1 1 100

245,290

1300

4320

793

2570

164,340

W m . Boby plant at Tobruk, N . Africa.

B.

S . O . D . E . M . I . installation, N . Africa.

to 500 p.p.m. NaCl (approximately 520 p.p.m. TDS) should not exceed R 0.25 per 1000 Imperial gallons (29 U . S. cents per 1000 U . S. gallons), including all pretreatment costs, and the complete plant and auxiliary facilities should be amortized over not more than 15 years. Design and cost studies indicated that, so far as could be seen at that stage of development, the desired low costs of production of desalted water would be achieved only with a plant having a small number of presses taking large membranes of the order of 6 feet by 2 feet 6 inches. In this way, if the engineering difficulties associated with such large units could be overcome, capital and operating costs should be at an acceptable level. The choice of size of unit was dependent not only on investment costs, but also on the availability of membranes of suitable size and price. Many other design and operating considerations had to be considered. These have been fully discussed (9). The type of membrane to be used was ultimately decided on the basis of cost, possible life, mechanical suitability, and suitable electrochemical properties. If the types of high quality membranes then available were to be used in units of the size considered desirable, a lifetime of at least 3 years was required to meet the desired desalted water cost, while for a plant with a larger number of relatively small units a considerably greater life would be required to break even on operating costs. There was, moreover, no guarantee that the numbers and sizes of high quality membranes required would be available with a guaranteed lifetime of 3 years or more. Under the circumstances, it was decided to base the design of the F S G plant on the use of SACSIR parchment-based membranes that had been used in the pilot plant. The estimated price of these was approximately 11 cents per sq. foot, with a 9-month proved life. Actual membrane prices for the plant were, in fact, almost double this figure, thus altering the price-lifetime relationship. The parchment-type membranes were originally developed for the pilot plant work as a stopgap, when the membranes ordered could not be delivered in time. As experience was gained with the membranes, it was considered that they might have an application in the desalting of low salinity brackish waters (4000 p.p.m.), and that they were particularly suitable for the mine water problem. In the latter case, their cheapness was attractive, particularly where process difficulties might require long-life membranes to be discarded in large numbers before their useful life was ended. The parchment membranes, in their improved form, gave very satisfactory service in the plant; if it had not been possible to use them,


ADVANCES

138

IN

CHEMISTRY SERIES

the F S G plant would probably not have been built, and no results on large plant operation would yet be available. After it had been confirmed that SACSIR parchment membranes could be made and safely handled in sizes up to 72 X 28 inches over-all, tests were carried out on a prototype unit (7, 9, I I ) , taking membranes of an over-all size of 66 X 26 / inches. Recalculation of the optimum operating conditions when using large membranes suggested the eventual size of unit taking membranes of 85 / X 25 / inches. This size was agreed on only after it had been confirmed that SACSIR-type parchment membranes of this larger size could also be made and safely handled. Means to ensure even flow distribution in the individual 22 / -inch-wide compartments had to be considered carefully. It is important to obtain rather uniform flow distribution over the whole face of the membranes, if polarizing conditions and scale formation are to be avoided, particularly in the case of the F S G plant, which was designed to operate originally with mean compartmental flow rates only some 10% above the critical polarization value. Experience in the pilot plant had shown that the ring distribution system from a single port (9) is not very effective with broad compartments. After a consider3

4

3

3

3


4

4

Figure 4.

4

Electrode and intermediate plates, gaskets, and spacer design


10. VOLCKMAN

Electrodialysis


Plant

139

able amount of experimentation on plates of the prototype size, the slot feeding distribution system (Figure 4) was decided on. It was always realized that there was danger of the membranes collapsing into the slots, particularly if pressures on the two sides of a membrane were considerably out of balance. Leakage from one compartment to the other could then take place (Figure 5). The dimensions of the slots were the largest that tests showed could be used without leakage. Unfortunately, these tests were carried out with fairly new membranes. In the plant the membranes gradually deformed with time, extensive leakage then taking place.


Operating Experience The operation of the F S G plant for over 18 months has given the first opportunity for assessing the characteristics of a fully integrated plant designed for multimillion gallons per day output, using a sheet-flow type of electrodialysis apparatus.

Figure 5.

Mechanism of slot or intercompartmental leakage

Membrane 1, having pressure on top greater than pressure between membranes 1 and 2, is pressed down into slot at A at edge of gasket. Slot is in gasket Y, feeding compartment between membranes 1 and 2. Deformation at A progresses until liquid in compartment on top of membrane 1 can pass under gasket X into conduit feeding short slot as well as into the correct slot feeding long conduit. Spacer omitted for clarity



140

ADVANCES

IN CHEMISTRY SERIES

The plant underwent its first tests in November 1958, with one press on line. Commissioning tests continued until April 1959. After the usual troubles experienced in starting up a new plant had been rectified as far as possible, the plant was put into routine operation under mine staff control in May 1959, with four of the eight presses on line. Delays in getting all eight presses into service were caused by very late delivery of some of the press intermediate plates. In June 1959, the mining companies decided to operate the plant on a production basis to obtain income from the sale of water, although at this time the plant had not been brought into full production. Except for special tests in April 1960, this period lasted until October 1960, when a series of investigations was carried out to confirm certain ideas as to why the plant was not achieving its designed output. From March 1960 to February 1961, a press with a single pack identical to those on the F S G plant was available at the SACSIR laboratories at Pretoria to check results obtained on the F S G plant, and to try out improvements in the design of pack components and methods of operation (10, 11). The unit was also used to test, under carefully controlled conditions, packs sent from the F S G plant. The only difference between operation on the units at Pretoria and at the F S G plant was that the former was fed with synthetic brackish water from which the clay particles found in the F S G raw water were absent.

T a b l e II.

Period 1959 3/29-4/25 5/29-6/25 6/26-7/25 7/26-8/25 8/26-9/25 9/26-10/25 10/26-11/25 11/26-12/25 1960 12/26-1/25 1/26^-2/24 2/25-3/24 3/25-4/24 4/25-5/24

Operating Time Actual, hr. hr.

Units Operating

%

Stage

Production Rate Mean U. S. g.p.h.

%•

590 259 .5 356 .8 631 .8 585 .5 714 661 .7 627 .2 644.2

87 .8 83 .2 99 .1 87. .8 78 .7 96 .0 91 .9 84 .3 95, .9

1 1 1 1 2 2 3 3 3

8,338 4,323 13,688* 13,040 30,080 35,300 40,960 45,000 43,954

27.8 14.4 45.7 43.4 50.0 58.8 45.5 49.9 48.8

637. .7 744 645 .8 715 541 561 438, .7 442 226 ) 201. 5 (6 units) j • 178 192 (8 units) j 655. .4 744 687. .8 720 663. .3 744 647 744

85. .7 90 .3 96 .4 99, .3

3 3 3 3

38,971 48,153 53,420 64,260

43.3 54.6 59.4 71.4

89. 2

3

51,766

57.51

92, .7

4

60,178

50.1

88. ,1 95. ,5 89. 2 87

4 4 4 4

56,940 56,584 61,120 52,347

47.4 47.1 50.9 43.8

672 312 360 720 744 744 720 744 672

j

5/25-6/24 6/25-7/24 7/25-8/24 8/25-9/22

D a t a f r o m Routine O p e r a t i o n

c

c

1

° % of designed capacity for section of plant actually in operation. Per electrode pair. Single diluting stream inlet connections to membrane packs. 6

c



10. VOLCKMAN

Electrodialysis


Plant

141

The successful technical operation of an electrodialysis plant depends, among other things, upon the production of components to very close tolerances, precise assembly, ease of handling and assembly of components, mechanical reliability, freedom from corrosion, low electrical losses, freedom from liquid leakage from concentrating to diluting stream or vice versa, freedom from external liquid leakage, freedom from scaling, and safety of operation. In the F S G plant difficulties were experienced in obtaining components having the necessary tolerances and intermediate pack plates having adequate mechanical strength and resistance to deterioration. Interstream leakage resulted in low electrical efficiencies and loss of output. Deposition of a clay sludge on the membranes proved troublesome; this was a problem peculiar to the particular mine water, but does emphasize the sensitivity of the electrodialysis process to the various substances that can be found in raw brackish waters, in particular membrane "poisons," scale-forming compounds, and suspended solids that can only with difficulty be coagulated or separated by filtration. In the latter case, the selfflushing characteristics of a particular design of plant can be important. In spite of difficulties, the F S G plant has shown conclusively that large sheet flow-type units can be assembled and handled without difficulty, even when using relatively weak parchment-base membranes. Operating reliability of the whole plant was of a high order in spite of defects developing in some of the pack plates. On-stream time over a continuous period

of Free State G e d u l d Electrodialysis Plant Kw.-HrJ 1000 U.S. Gal. Product 1000 Coulomb Efficiency, % Mean Electrical Conditions p.p.m. Max. Mean Amp. Volts at 25° C. Mean TDS Stage Stage Stage Stage Temp., Stage Stag" Stage Stage removed, II 25' C. II I C. II II I 25° C. 1

19! 8 22. 8 25. .6 23. .8 25. .9 28. 1

70 69 65 68 72 70 70 65 59

76 51 79 74 74 75 74 70 67

80 83 77 81 86 84 83 73 63

90 63 85 89 83 86 89 81 80

14.6 18.7 9.5 9.7 9.6 10.2 11.7 12.7 13.0

4. .93 5..26 4. .60 5 .20 5 .95 6 .33 6 .55

34. .7 33. .3 34. .5 36. .4

27. .3 27. 3 27. 3 25. 8

60 75 76 78

66 69 75 72

65 92 87 94

86 86 91 82

13. 8 11. 6 11. ,1 7. .75

7.33 5.61 5.13 3.82

58 .8

38 .7

21. .6

71

73

74

75

8. .6

4.45

55, .8 55 .8 53 .0 53 .6

32 29 28 37

.9 .7 .2 .0

19. .9 19 .1 22 .4 24 .0

70 76 69 65

76 77 71 73

79 91 79 80

84 94 89 94

9 .4 8 .2 8 .75 9 .3

4.86 4.56 4.75 5.11

782 784 800 778 831 850 837 880. 6 885. .2

579 587 558 565 603 646 687 665.3 687.1

41 33 63 60 67 79 70 76.5 75.8

23 24 26 27 30 35 32 38. 6 37. ,5

23 .0 20. 9

792. 9 856. 1 869. .7 842. 5

635. 4 644 654. ,3 622. ,7

86. 5 83. .5 82. .3 68. .7

782. .9

572. .6

753, .1 742 .1 792 .7 837

573, .7 548 .6 594 .4 635 .1

Double diluting stream inlet connections to membrane packs for this and all subsequent periods of operation. d


142

ADVANCES

IN

CHEMISTRY SERIES

of 3 months was 93.6%, while for some individual months close on 100% was achieved (Table II). Electrical short-circuiting was less than 2.5% of the maximum current; the plant was safely operated at voltages of 450 to ground.


The electrical resistance of the units in stage 1 increased by about 100% over a period of 10 to 15 days. This was due almost entirely to accumulation in the diluting compartments of a clay sludge from colloidal clay which was present in the treated raw water in amounts varying from 0.05 to 0.3 p.p.m., with an average of 0.25 p.p.m. The complete removal of the colloidal clay particles found both in these mine waters and in surface waters over a large area in the region is a considerable watertreatment problem. Whereas residual suspended clay particles totaling 1.0 p.p.m. in a treated water will give domestic water of reasonable clarity, such amounts are unacceptable in the electrodialysis plant. Amounts not exceeding 0.05 p.p.m. are preferred, but this requires exceptionally good water treatment control and the use of a "polishing filter" of the precoat type in the final stage. No polishing filter was used in the F S G plant. To remove the accumulated sludge, the packs had to be scoured with a flush of compressed air and water every 7 to 10 days. The electrical resistance of the units was reduced to the original value in this way. Membranes and graphite electrodes gave a little longer than the estimated service of 9 and 6 months' continuous use, respectively. A qualification is needed in the case of electrodes, as the plant operated for much of its time at well below the maximum designed current density, because the ohmic resistance of the packs, due to sludge formation and polarization, rose to approximately twice the designed value. Labor requirements on the F S G plant were normally one operator per shift to check plant operation, plus three semiskilled operators employed for membrane pack assembly and overhaul, and general duties resulting from operating the plant partly on a development basis. On a routine basis two semiskilled operators on day work, instead of three, would be adequate. The plant was large enough to justify a full-time plant superintendent. Varying numbers of Africans were used for cleaning and miscellaneous unskilled duties. Engineering and instrument maintenance was on a contract basis from the mine services. For pack assembly and membrane replacement a minimum of two operators on day work was required. The operators were also required for unpacking, washing, trimming, and punching membranes, which were supplied to the plant as plain sheets slightly oversize. They also assisted the plant operator when a press was started up or shut down, and with filter plant and pretreatment plant operation, although this was not essential. The human element in assembling packs should not be overlooked. This is particularly important in large plants, and could become an increasing problem the greater the number of packs and components in a plant. Thus, although there may be an argument in favor of using more cell pairs each of individually smaller area, instead of fewer units of larger individual membrane area, it must be realized that assembling membrane stacks is extremely monotonous, but nevertheless requires a certain degree of skill and attention to detail. In the F S G plant, pack overhaul was determined largely by the life of the positive, or anion-selective, membranes. Their life was approximately 10 months, but for calculation purposes 9 months was allowed for. On this basis, approxi-


10. VOLCKMAN

Electrodialysis


Plant

143

mately 110 membrane packs had to be opened up and serviced each year. In practice, the number could be slightly larger to allow for possible early failures. Ideally, a plant of this size is started up in stages, so that replacement will be a continuous process. Membranes in all the 10 packs of a press are generally renewed at the same time. A reserve of 10 assembled packs can be kept in a water tank for immediate installation. If necessary, a single pack in a press can be renewed at any time, although this complicates the maintenance schedule. The reserve press on the plant, which could serve either stage I or stage II, was used when a press was being serviced. On the basis of servicing 110 packs per year, approximately two packs a week would have to be repacked. Experience in the F S G plant showed the following schedule to be a conservative one:


Schedule" 1. 2.

Shut down press and drain packs Remove hose connections and position pack clamps Remove packs Reposition new packs Align packs and close press Replace hose connections and startup

3. 4. 5. 6.

Replacement of All 70 Packs in Press at One Time, Min.

Replacement of Single Pack Only in Press, Min.

15

15

80 60 60 45

80 10 10 15

150 6.8 hours

150 4.7 hours

Schedules 2 and 6 could be considerably reduced for single-pack replacement if more flexible water-inlet hoses were fitted. The hoses used were rather inflexible and had to be removed before a press was opened. Pack dismantling, checking of components, and reassembly took approximately 10 hours per pack. This time could be improved, but could not be maintained week after week while still maintaining the required standard of assembly. a

The two pack assembly workers would therefore be occupied on day work throughout the year thus: 12 Presses to be serviced per annum, 8 X Packs to be serviced per annum Hours per annum for servicing presses (allowing 1 day per press), say 11 X 8 Hours per annum for servicing packs, 1 1 0 X 1 0 Based on 40-hour week

11 minimum 110 minimum 88 1100 1188 29 .7

minimum minimum hours weeks

Allowing for contingencies, approximately 32 weeks per annum would be required for press replacement and pack overhaul. The balance of 20 weeks is taken up by unpacking, soaking, trimming, and punching membranes, assisting with pretreatment plant operation, miscellaneous maintenance, statutory holidays, and leave. These other duties are highly desirable to relieve the workers of the monotony of pack assembly and to prevent mistakes that can be caused by it. It is recommended that in any large plant using multipack presses, facilities be available for hydraulically and electrically testing packs after assembly and before installation in the presses. Where very large plants are under consideration, the amount of labor required for membrane pack maintenance should be carefully examined. This will depend on membrane life, or other factors requiring packs to be dismantled, such as scale damage or membrane poisoning, the number and size of units employed, and the time required for removing the packs from the units and servicing them. Integral membrane-spacer-gaskets were being developed by the SACSIR in



144

ADVANCES

IN

CHEMISTRY SERIES

collaboration with an industrial firm, but this work, although promising, has been discontinued. Such spacer-gaskets might decrease handling time, but it is doubtful if any real advantage would accrue, as pack assembly workers would, in any case, have to be employed on a full-time basis. Damage to a gasket would also mean rejection of a spacer, and vice versa. Apart from difficulties inherent in the scale-up from pilot plant to prototype, other difficulties arose when transferring from prototype tests on single-pack presses to multiple-pack presses. Thus, while single packs could be easily sealed, considerable difficulty was experienced in satisfactorily sealing the packs of the F S G presses. The friction between the gaskets and the membranes was relatively low, so that if the water pressure in a compartment was too high the gaskets tended to bow out and form a low resistance flow path between the edges of the gasket and spacer. This caused a lower than average water flow rate over the main area of the membrane, aggravating the risk of polarization in diluting compartments. In single-pack tests a small increase in sealing load can rectify leakage, whereas in large multiple-pack presses a considerable amount of friction is encountered when the packs move as they are squeezed up, so that a much larger force has to be applied to seal the unit. From prototype tests it seemed that a net sealing load on the gaskets of as low as 30 p.s.i. would be adequate. In practice, the sealing load that had to be applied was 120 p.s.i., calculated on gasket area. The sealing load affects not only the amount of external leakage, but also the degree of internal leakage. If the pressure drop across a compartment (measured at the inlet and outlet manifolds) rises to 20 p.s.i., the minimum sealing load required on a single-pack assembly, calculated on gasket area, is 80 p.s.i. This is sufficient to ensure no external leakage, and, with adequate support for the membranes at the gasket slots, internal leakage can be kept down to approximately 0.2% of full flow.

Plant Operating Data The operating history of the plant is given in Table II. At the start of the period shown the units on line had already operated for several months during preliminary testing. From the beginning of February 1960 to the end of April 1960, membranes were gradually replaced in the units that had been on line for 9 months and more; the resulting progressive reduction in energy consumption per 1000 p.p.m. T D S removed is seen in the last column. [These values are given for comparison purposes, as the degree of desalting varied from time to time. In assessing the data presented it is important to realize that, unlike bore hole waters, there were wide variations in raw water salinity (depending on from which shafts water was being pumped) and from dilution by surface waters entering the holding dams during storms. Input concentration varied from 2780 p.p.m. maximum to 1954 p.p.m. minimum mean daily values.] The production rate was below design because of intercompartmental leakage, and the desalting below design because of high cell electrical resistance due to excessive polarization. The current shown is the maximum that could be supplied at the maximum voltages obtainable from the rectifiers.

Capital Costs Estimated and actual capital costs are given in Table III.


10. VOLCKMAN T a b l e III.

Electrodialysis


145

C a p i t a l Costs f o r Free State G e d u l d Electrodialysis Plant

Pretreatment plant Plant items, pumps, lime and acid-dosing equipment, sand filters, etc. Buildings Piping, valves, and fittings Instrumentation Erection and site preparation Contingencies and overheads Subtotal


Plant

Estimate {1957),% 73,500 65,700

&

Actual* (1959),% 4O,950 19,800 29,300 60,050

c

c

d

6,380 7,270 152,850

150,100

179,000

282,000

Electrodialysis plant Electrodialysis units, excluding piping and instrumentation, but including first charge of membranes and electrodes Other items, H . T . and L . T . transformers, switch gear, rectifiers, pumps, storage reservoirs, etc. Building, including laboratory and membrane pack assembly area (store and workshop under filter plant) Piping, valves, and fittings Instrumentation Erection and site preparation Contingencies, overheads, and royalties Subtotal

276,400

191,000

102,800 35,650 19,600 67,500 73,450 754,400

39,400 69,700 61,900 3,500 647,500

Total

907,250

797,600

c c d

Prices converted to U . S. dollars at rate of R1.00 = $1.40. ° Values rounded off to nearest $50. Including piping, instrumentation, etc. Including erection. Excluding erection.

b c

d

Production Costs Table IV gives actual and estimated production costs for a number of conditions. Costs cannot be directly compared, as they are affected by plant output and the degree of desalting. In this particular case they can, however, be compared more nearly if calculated per 1000 gallons and per 1000 p.p.m. T D S removed. The costs in column 2 are based on operation for the month of April 1960, in so far as output, degree of desalting, and chemical and electricity consumption are concerned. The other costs are based on average as accurately as could be ascertained. Averaging over a 15-year life, repairs and maintenance might be slightly higher. A membrane life of 9 months has been proved. No estimate of the ultimate life of the gaskets and spacers was possible. Experience indicated that they could be expected to last for many years in service. The allowances made in the estimates should be achieved in routine operation. Column 3 of Table IV is extrapolated from the actual operating data used for column 2, for the plant operating with the fourth set of presses put in line, but still operating below design for the reasons given above. Column 4 represents costs based on plant modifications enabling polarization and intercompartmental leakage to be eliminated, and full capacity and desalting to be achieved. The modifications would have involved additional work to the extent of $142,900. This cost has been added to the original costs in calculating amortization. No allowance has been made for credit for redundant equipment, or for the fact that these modifications if originally incorporated would actually have reduced the capital costs. Replacement and service costs are based on conservative values, as shown in the notes to the table. A mean coulomb efficiency of 75% throughout the plant has been allowed for, with membranes discarded


146

ADVANCES

T a b l e IV.

Production Costs f o r Desalted W a t e r f r o m Free State G e d u l d Electrodialysis Plant Modified F.S.G. Plant Operating at j 700% 77.4% Capacity per Unit, 7960 Capacity, Actual Calcd. Calcd., 1967 3 4 4 64,260 85,500 120,000 2,843 2,843 2,900° 525 813 813 or

Estimate, 7957 Units operating per stage Desalted water, U.S.g.p.h. Raw water concn., p.p.m. T D S Desalted water, p.p.m. T D S


A.

Pretreatment Chemicals Electricity Labor Repairs and maintenance Interest and redemption^ Subtotal

B.

IN CHEMISTRY SERIES

Electrodialysis Replacements Membranes Electrodes Miscellaneous* Electricity Electrodialysis' Pumping, etc. Labor and administration Repairs and maintenance Royalties Interest and redemption^ Subtotal Total cost Per 1000 U . S. gal. produced Per 1000 U . S. gal. Per 1000 p.p.m. T D S removed % T D S removed p

4 120,000 3,100 525

Costs in U . S. Cents/1,000 U . S. Gal. Produced 0.45 1.35 1.00 0.80 Included in B 0.13 Included in B 0 .41 Included in B 1.47 2.06 2 .74 1 .54 2 .53

6

c

3 .66" 0 .29* 1 .41

10. 22' 0 .20' 2 .63

10.22'.* 0.20' 1.97

6.85'.* 0.14'" 1.41

4.48" 1.289 3.769 1.439.' 0.42 8.91

5.25° 1.189 3.229 1.029.'0.30 8.04

5. 89 0 .49 4 .38 1 .19 0 .51 7 .31 25. .13

4 .48" 1 .70? 5 .00? 1 .919,' 0 .56 11 .86

27, .7

42, .65

35.73

29.7

10, .7 83, .3

21, .0 71. .4

17.6 71.4

11.9 82.1

m

° 2900 T D S was absolute maximum value expected from any mines serving plant at this date. Costs converted to U . S. currency at R1.00 (S.A) ± $1.40 ( U . S.). 0.58 cent/kw.-hr. Redemption over 15 years; interest on capital 6.5% per annum payable quarterly = 10.3% per annum. For capital costs see Table III. S A C S I R parchment membranes at $1.75 each delivered; 9-month life; 21,330 membranes per annum. ' $3.60 each delivered, including royalty; 9-month life; 16,000 membranes per annum. 21,330 membranes per annum. 8-month life and 16,000 membranes per annum. Graphite, 3-month life. ' Graphite, 6-month life. * Including sundry spacer and gasket replacements, estimated. A.c. energy. A.c. energy at 0.58 cent/kw.-hr., 90% a.c./d.c. rectification efficiency. A.c. energy at 0.52 cent/kw.-hr., 90% a.c./d.c. rectification efficiency; 8.60 kw-hr. a.c. per 1000 TJ. S. gallons produced. ° A.c. energy at 0.52 cent/kw.-hr., 92% average a.c./d.c. rectification efficiency at higher loads possible. 10.1 kw-hr. a.c. per 1000 U . S. gallons produced at mean annual conditions of water temperature, membrane properties, and other plant operating conditions. P A.c. energy at same prices as for electrodialysis; includes lighting and other miscellaneous services. « Includes pretreatment plant. Supervisor R240.00 per month basic salary; chemist Rl67.00 per month; shift operator R3.97 per shift; packer R3.64 per shift; African laborer R20.00 per month, net. 6

c

d

e

0

h 1

1

m n

r

when giving 65% instead of 60% coulomb efficiency, thus reducing membrane life from 9 to 8 months. This is achieved by a planned replacement of membranes in


10. VOLCKMAN Table V.

Electrodialysis


Estimated Costs f o r Plant Based o n O S W S t a n d a r d i z e d Estimating Procedure (3) A.

Capital costs

0

$ 254,000 84,000 338,000 101,100 13,500 452,600 18,000 137,400 28,820 6,710 643,530 64,200 707,730 72,400 31,200 8,640 819,970 65,100 885,070

Special equipment (installed) Standard equipment (installed) Total PIE* (installed) Erection and assembly Instruments Total essential plant costs Raw water supply Extra facilities Product water storage Service facilities and buildings Contingencies Engineering Interest on investment during construction Site Total plant investment Working capital Total capital costs


147

Plant

Cost per U . S. gallon per day of production B.

Operating costs (Basis.

c c

0.307

1 stream day and 330 operating days per annum)

Essential operating costs

Cost per Stream Day, $ I II*

Electric power Chemicals Membranes Electrodes Supplies and maintenance Operating labor Maintenance labor Payroll extras Total essential operating costs

224.50 12.88 105.80 8.40 12.31 224.00 12.31 35.22 635.42

Other operating costs General overhead and administration Amortization Taxes and insurance Interest on working capital Total operating costs for one stream day Cost per 1000 U . S. gallons of product Approx. cost per 1000 U . S. gallons of product per 1000 p.p.m. T D S removed

d

c

81.50 183.30 49.20 7.00

c

c

214.00 23.01 197.20 4.04 13.70 224.00 35.65 38.94 750.54

c

89.58 205.00 54.90 7.98

c

c

956.42 0.332

1108.00 0.384

0.133

0.160

° For plant as designed and erected. Actual prices used where available. Principal items of equipment. Not included in S A C S I R estimates. Based on 1957 estimate values. Based on conditions applying to Table IV, col. 4.

b c

d e

the presses.

T h e values allowed for coulomb efficiency are probably conservative,

as efficiencies of membranes f r o m the plant after 10 months' use still averaged 7 0 % w h e n used i n a small laboratory electrodialysis cell. T h e plant modification for costs i n column 4 includes e q u i p p i n g each press w i t h its o w n diluting and concentrating stream pumps, a l l o w i n g for equal flow rates w i t h 4.1 to 1 concentrating stream recycle, compartment flow rates of 40 U . S.


148

ADVANCES

IN

CHEMISTRY SERIES

gallons per hour instead of 30, and 75 compartment pairs per pack instead of 100. Other conditions remain as before. No allowance is made in the costs for taxation, cost of site, insurance, etc., as circumstances vary so greatly. Estimated capital and operating costs, based on the Office of Saline Water standard procedure for cost estimating (3), are given in Table V . Agreement with Tables III and IV is good. Estimates for a plant operating under conditions applying to column 4 of Table IV, but with very much simpler piping and control arrangements, would be possible only if plant operational data over a much greater time had been available. It would be premature to present an estimate at this stage, particularly as the question of plate fabrication has not been satisfactorily resolved.


Reduced Plant Capacity As will be seen from Table II, the F S G plant did not operate much above about 55% of design output for any length of time, 71% being the highest value obtained and this only at the expense of some reduction in the degree of desalting. The reasons for this reduced output are important from a design point of view. 1. The factor preventing the designed output from being obtained was the onset of polarizing conditions in the diluting compartments at average nominal linear liquid velocities across the compartments higher than expected from the pilot plant and prototype work. The development of polarizing conditions was aggravated by the use of unequal flow rates in the diluting and concentrating compartments. This caused opening of the diluting compartments in spite of out-of-balance pressures being minimized by the application of a back-pressure on the concentrating compartments. Linear flow velocity in the diluting compartments was thus Tower than that calculated for the compartment of designed thickness. A change in the intermembrane spacer corrugation angle used also contributed to a higher critical polarization velocity. The sensitiveness of compartments to change in thickness from out-of-balance hydraulic pressures in adjacent compartments was much greater with the large areas of the F S G plant than with the pilot plant, where the membrane pack assembly was relatively rigid. The polarization difficulties were overcome by using symmetrical flow systems and flow rates in both diluting and concentrating streams. 2. A contributory cause was the considerable leakage of the diluting stream into the concentrating stream, referred to as "intercompartmental leakage." This leakage was eventually held to 0.3% of the diluting stream flow, but before modifications were introduced (aimed at supporting the membrane at the feed slots in the gaskets), values were at times as high as 25%. 3. Another difficulty was a mechanical one regarding materials of construction for the pack, electrode, and intermediate plates, which were all of the same general design, differing only in details of liquid ports and conduits, etc. In particular, leakage from one stream to the other occurred in many of the timber plates because of opening up of the glued joints; this permitted interconnection from a main diluting stream conduit to a main concentrating stream conduit. 4. Some trouble was also experienced with the welds of 8-inch diameter highdensity polyethylene pipe, which was later replaced by rubber-lined mild steel, or acrylonitrile-butadiene copolymers. The latter was used for all the smaller diameter piping, but was not available in the larger sizes (8-inch diameter and above), when the plant was being built. Most of these difficulties occurred in transferring from pilot scale to full plant scale with only limited prototype work. The reasons for this have been stated above. Saline Water Conversion—II Advances in Chemistry; American Chemical Society: Washington, DC, 1963.

10. VOLCKMAN

Electrodialysis

Water-Deminerallzation

Plant

149

Electrodialysis Press Components and Effect on Operation


Intermediate and Electrode Pack Plates. The construction of reliable intermediate plates and electrode plates proved difficult. The general design of the plates is shown in Figure 4. Membrane support grids of design other than the form shown were also used. The prototype plates were made from glass fiber-reinforced polyester resin and those on the pilot plant from glass fiber-reinforced epoxy resin. Although some distortion, surface rippling, and blow holes occurred on the prototype plates, the manufacturers were confident that the problems could be overcome and that rejects would be few. Just before the order for plates was due to be placed, the firm discontinued work in this field and the whole question of materials was thrown back to the beginning. It is only recently that satisfactory molding or casting techniques have been developed which might enable glass fiber-reinforced resin sections of the size required on the F S G plant to be reliably made. Even so, sections of the size and thickness required have not yet been made. When the plant was being designed, typical alternative materials were rigid poly (vinyl chloride), rubber-lined steel, plastic-coated alloys, and timber. Rigid P V C was excluded on a cost basis; as it was available only in sections up to 1 inch thick, lamination would have been required and labor costs would have been considerable. Even so, as events have turned out, this might have ultimately been the cheaper solution. Rubber-lined steel was eliminated because of weight, extra size, and expense, and plastic-coated materials because of possible damage to thin plastic coatings and the, then, limited number of reliable coating techniques available in South Africa. Work in Pretoria had shown that hardwoods like African kiaat and phenolformaldehyde resin-impregnated European beech, glued with phenol-formaldehyde-resorcinol glues, could be expected to give reasonable service, and that weight and prices were reasonable. The plant was equipped mainly with beech plates impregnated with phenolformaldehyde and glued with a phenol-formaldehyde-resorcinol adhesive. Experience showed that unless extra special care is taken in selecting, curing, impregnating, and working the timber, reliable plates cannot be produced. Laminated beechwood plates made from material extensively used for electrical insulation in submerged conditions have shown no better results. Extensive warping and splitting of the material at the joints between laminates occurred and caused considerable difficulty in the plant. A few electrode plates were made of Perspex (Lucite). These suffered from incipient cracking, and, in any case, were expensive. They were used only to fill a gap before the main order for timber plates was completed. Although kiaat plates gave good service over a long period of testing on the unit at Pretoria and were repeatedly dried out, timber plates cannot be recommended for mass production and plant use. Experimental full-size plates of glass fiber-reinforced polyester resin, and also of a P V C or polyethylene-coated aluminum alloy have been made and tested. The plastic coating of the alloy plates was reparable on site. It was electrically tested for pinholes and was tested for abrasion, tear resistance, and resistance to stripping. The glass fiber-reinforced polyester plates needed considerable macriining before being put into service. The plastic-coated aluminum alloy plates are the cheapest, but there is always danger of undetected damage of the lining.


150

ADVANCES

IN

CHEMISTRY SERIES


Comparative costs for quantity production are approximately in the ratio of 1 to 4 for the aluminum alloy and the polyester plates, respectively. In designing a large plant a balance must be struck among membrane, and hence plate, dimensions, available materials of construction, and their cost and workability. If fabrication costs are not important, rigid P V C is suitable. On the thick sections now available considerable machining is required. Glass fiberreinforced resin plates are limited to sizes determined by production techniques. Plastic-coated cast metal plates are relatively cheap and lend themselves to large sizes, but suffer the disadvantage of all coated materials, that the coating may be damaged and the metal may then corrode. Timber does not seem a satisfactory material, even if impregnated with modern resins and bonded with modern waterproof marine glues. From Figure 4 it will be seen that in the intermediate plates there were grids to support the membranes and yet permit electrical continuity. In starting or stopping the plant there was always the chance that the intermediate plate rinse stream might be started up after, or turned off before, the streams through the packs. If this happened, pressures up to 20 p.s.i., depending on outlet piping back pressures, could be exerted on the grids by expansion of the membrane pack. This had, on occasions, happened and been sufficient to break the grids and to damage the end membranes of a pack. In the earlier plates breakage of the side stiles occurred due to the pressure of the rinse stream in the plates. Later modifications included insulated tie rods. Grids integral with the plate, which were used to strengthen the side frames of timber or Lucite plate, were also used. In this latter case, care had to be taken when building the packs that the tolerances of the components were such that the finished pack was substantially flat, or only slightly bowed outward. Excessive bulging could lead to breakage of the grids when the press was tightened up. Where freely supported grids were fitted, a type made from strips of corrugated glass fiber P V C sheets placed end-on and solvent-welded to each other to form a honeycomb, was promising. It had considerable strength against deformation at right angles to its length or breadth. Gaskets and Membrane Spacers. In considering tolerances of components, reference must be made to the gaskets and spacer material used in the F S G plant. In all the early work it was considered advisable to use a gasket that was fairly rigid, would not suffer cold flow, would seal against the membranes at a reasonably low sealing pressure, and could be re-used repeatedly. Klingerit, a high quality steam jointing, fulfilled these requirements and also had a high degree of dimensional uniformity in thickness. The same degree of uniformity was not achieved, however, with the corrugated perforated spacer material, the manufacturer averaging not better than ± 0 . 0 0 4 inch on the stipulated thickness of 0.036 inch. Tolerances were fixed at ± 0 . 0 0 2 inch. This led to difficulty in assembling packs—the use of too much oversize spacer made it difficult to close the pack, while the use of too much undersize spacer led to inadequate support of the membranes. It is now considered that rather than trying to match the spacer thickness to the gasket, the reverse is to be preferred, as with a somewhat resilient gasket variations in spacer thickness can be accommodated and a firmer membrane assembly be obtained. One reason for using more or less rigid gaskets of a uniform thickness was the desire to obtain a uniform mass flow rate in all the compartments of one type, particularly since the designed flow rate in the diluting compartments was fixed



10. VOLCKMAN

Electrodialysis


Plant

151

rather closely to the critical polarization value. This feature, and the use of narrow compartments, were the result of having to design for o p t i m u m conditions. Critical Polarization Velocities, Flow Distribution, and Pack Electrical Resistance. The critical polarization velocities allowed for were based on results from average packs i n the pilot plant and some confirmatory testing on the prototype. It was considered that this w o u l d allow for normal flow variations from pack to pack and from compartment to compartment. A method used i n this laboratory to obtain a picture of the flow distribution in a pack (11) is based on the increase i n resistance occurring as polarizing layers b u i l d up, and is a function of l i q u i d velocity. In practice, platinum wires are inserted through the gaskets into the compartments and the voltage drop between a number of compartments is measured. T h e method can be extended to determine conditions i n any portion of a compartment. Although actual velocities cannot easily be calculated from the data obtained, a general picture of flow distribution i n a pack is given. Absolute flow rates and velocities relevant to the observed voltage drop can be obtained from studies on a single compartment, or a single cell pair. Figure 6 shows typical data so obtained for standard F S G - t y p e packs. It shows also how the pack electrical resistance, as a whole, decreases as the flow rate is increased, the general flow distribution pattern remaining. E q u a l i t y of distribution is dependent not only on uniformity of the components and accuracy in forming the conduits, but also on the manner in w h i c h the streams are introduced and taken from the packs—e.g., at one or both sides of a plate, or at one or both plates, and variations of this (10, 11). Critical polarization velocities were found to be some 2 5 % higher than expected i n the F S G plant, resulting i n a pack resistance from 1.8 to 2.3 times higher than designed for at standard conditions of flow, concentration, and temperature ( 2 5 ° C ) . This was eventually traced to the configuration of the spacer material used. In the pilot plant maximum turbulence was desired i n the diluting compartments, i n which the corrugations of the spacer were placed at right angles to the direction of l i q u i d flow (in the concentrating compartments they were placed parallel to the direction of flow). In the F S G plant the spacer material corrugations were arranged to be symmetrical i n the compartments, so that identical conditions w o u l d exist when polarity was reversed. It was not possible to place the corrugations at right angles to the direction of l i q u i d flow i n every compartment, as the membranes might i n time have deformed into the corrugations. Preliminary tests w i t h the corrugations set at 8 5 ° resulted i n such a high back-pressure i n the diluting stream relative to that i n the concentrating stream, where the flow was one quarter of the diluting stream, that the membranes deformed into the grooves of the spacers. Corrugations were therefore set at 15° to the direction of flow (see F i g u r e 4 ) . This change in angle raised the critical polarization velocity to the extent mentioned above, but the effect was not observed on the prototype tests, possibly because, although the same angle was used, the material lacked uniformity. F o r a given set of conditions the average ohmic resistance of a pack, expressed per cell pair, increased as the number of compartments increased, indicating possibly that w i t h packs of large numbers of membranes there is considerably greater chance that one or more compartments w i l l have a l i q u i d velocity below the critical. This may, in part, be due to the inherent flexibility of large packs with large numbers of membranes. These observations led to the necessity of increasing the flow rate per diluting



152

ADVANCES

IN

CHEMISTRY SERIES

A Total flow to all comportments.(Equal diluting and concentrating streom flows.) 3000 U.S. gp.h. I 0

20

Figure 6.

AO

60

80

3750 U.S.gp.h. j i 100 0 20 40 60 80 100 0 Comportment Number.

3960 U-S.q.pth. J L 20 AO 60 80 100

Distribution of electrical resistance across 100-compartment pack, and effect of over-all flow rate

compartment in the packs of the F S G plant from 25 Imperial gallons per hour (30 U . S. gallons per hour) to 33 Imperial gallons per hour (40 U . S. gallons per hour). Intercompartmental Leakage. Intercompartmental leakage occurred by leakage across membranes behind the feed and outlet slots in the gaskets (see Figure 5). The plant was designed for a 4 to 1 ratio of flow rates across diluting and concentrating compartments, approximately equal pressures in the compartments being achieved by throttling the concentrating stream outlet to achieve a backpressure at the pack inlet equal to that at the diluting stream inlet. Absolute equality of pressure is not possible, as is seen in Figure 7. This represents the approximate relative pressure conditions at designed flow rates. The slot feeding arrangement shown was chosen partly to balance pressure losses in adjacent compartments, and partly because the use of reversed polarity for scale control meant that the diluting compartments were used as concentrating compartments when polarity was reversed. There is some doubt if the reversed polarity procedure on this plant was



10. VOLCKMAN

Electrodialysis


Plant

153

justified. Normally, reversal was done daily, but scaling was not particularly serious, and the troublesome barium-strontium sulfate scale obtained from the mine water used in the pilot work was absent. Acid dosing of the brine to p H 4.5 to 5.0 controlled scale formation from calcium and magnesium carbonates in the water. Elimination of polarity reversal would obviate the switching and valve change-over arrangements, with the necessity for sizing all valves and piping for the diluting stream flow rates. This last point falls away, however, if the principle of equal flow rates with brine recycle is adopted. The possibility of "slot leakage" was considered very carefully and investigated on the prototype unit. The dimensions of the slots used in F S G plant were the result of extended laboratory tests to check that the membrane would not collapse into the slot. Prolonged operation proved, however, that over a period of time the negative membranes deformed into the slots; the positive membranes also did this, but to a much lesser extent. The flow of liquid across the surface of the gasket at the leakage point led to softening of the gasket at that point. Once leakage started, it got rapidly worse. Leakage seemed to be associated mostly with the negative membranes, which had a higher activity and degree of swelling. Whether or not this point is relevant has not been determined. The positive membranes showed a marked tendency to stick to the gaskets, and this sealing tendency may be of importance in preventing leakage. Pressures in P-s.i. o

o o o o o o

Uilliltt;

o-o 0-5

6-1

0-6

0-6

7-2 Diluting compartment.

Concentrating compartment.

Figure 7. Pressure drop across and pressure differential between diluting and concentrating compartments for 30 17. S. gallons per hour per compartment Absolute pressures in compartments are higher because of hydraulic resistance of outlet piping system Saline Water Conversion—II Advances in Chemistry; American Chemical Society: Washington, DC, 1963.

A D V A N C E S IN CHEMISTRY SERIES

154


Even when only small pressure differences exist between adjacent compartments, considerable leakage can occur i n packs of 100 membranes and over. Thus, with the slot feeding system as designed, a 25% difference of flow between adjacent compartments was sufficient to cause a difference of pressure at the inlet to the pack of 4 p.s.i., giving leakage of the order to 10% of the standard diluting stream flow rate. W i t h a 4 to 1 ratio of diluting stream to concentrating stream and balanced outlet pressures it was not possible to achieve less than a 0.6-p.s.i. pressure difference at the inlet, w h i c h , w i t h membranes that h a d been i n use for some time, allowed an uneconomic degree of leakage. By modifying the slots so that the membrane is supported (Figure 8 ) , leakage at substantially equal flow rates can be held at less than 0.3% of the diluting stream flow rate.

*- Alternatively corrugated spacer motgrial.

Figure 8.

Alternative gasket slot feeding arrangement for equalizing pressure drop and supporting membrane

Plant Capacity T o bring the plant to f u l l capacity w o u l d have required the elimination of polarizing conditions and of intercompartmental leakage. T h e former could have been achieved b y increasing the diluting stream flow to 133.33% of design a n d the latter by using equal diluting and concentrating stream flow rates and adequate membrane support i n the feed slots. O n increasing the diluting stream flow rate to 40 U . S. gallons per hour per compartment, the average pack electrical resistance fell to slightly below the designed values, thus enabling f u l l designed current densities to be used. T h e Saline Water Conversion—II Advances in Chemistry; American Chemical Society: Washington, DC, 1963.


70.

VOLCKMAN

Electrodialysis


Plant

155

degree of desalting per pass was, however, reduced in proportion to the increase in mass flow rate. Recalculation of the plant design for the new flow conditions showed that the originally designed stage conditions of current density should be retained. The required desalting could have been achieved at all times by allowing for a minimum average coulomb efficiency of 65% instead of 60%. This value would have been possible instead of the 75% average coulomb efficiency mentioned above and used for column 4 of Table IV, because the salinity of the mine water was still decreasing and over the last 4 months of plant operation had approached 2700 p.p.m. T D S . The total membrane area in the plant would be reduced to 75% of the original. Membrane life would be reduced from 300 days at a 60% minimum coulomb efficiency to 270 days at 65% efficiency. This is probably conservative, because plant membrane efficiencies as checked on small laboratory cells averaged 70% after 10 months' use. There is, thus, a net saving of approximately 17% on the annual requirement of membranes. An additional safety margin is that the rectifiers were designed for maximum voltages at a minimum water temperature of 1 5 ° C . In fact, on only one day had the water temperature fallen to 1 7 ° C. The mean monthly minimum water temperature was 1 9 . 1 ° C . For most of the year temperatures ranged between 2 4 ° and 2 8 ° C. The temperature rise of the water in the plant did not exceed 2 ° C . The use of equal flow rates would not have been a large complication, because most of the concentrating stream system was sized for the maximum flow rate of the diluting stream, to allow for reversed polarity. The only remaining factor requiring extensive testing in the plant would have been the reliability of the revised scheme for maintaining intercompartmental leakage at or below 0.2% of the diluting stream flow. External leakage could be kept well below 0.01% with plastic-covered plates or timber plates in good condition. The proposed revised plant conditions are shown in Table VI. T a b l e VI.

P r o p o s e d Revised O p e r a t i n g Conditions f o r Conditions Stated

Inlet water, U . S. gal./hr. Desalted water, U . S. gal./hr. Concentrated water, U . S. gal./hr. No. of operating presses No. of packs per press No. of membranes per pack

150,000 at 2700 p.p.m. T D S max. 120,000 at 525 p.p.m. T D S 29,400 at 15,800 p.p.m. T D S 8 (stage I, 4; stage II, 4) 10 150 Stage II

Stage I Current density (max.), ma./sq. cm. Desalted water, p.p.m. T D S Concentrated water, p.p.p. in T D S Concentrated water, p.p.m. out T D S Flow rate per compartment, U . S. gal./hr. Diluting stream Concentrating stream Recycle on concentrating stream

1

16 1,190 14,400 15,800

520 6360 7100

40 40

40 40

4.1:1

Membranes Average values of membranes used are given in Table VII. The values for discarded membranes represent those for the original installed membranes, which were of appreciably lower quality than the ones later in use.


156

ADVANCES T a b l e VII.

A v e r a g e M e m b r a n e Properties Cationic

Free electrolyte diffusion at 30 ° C . between distilled water and 15,000 p.p.m. N a C l , meq./sq. cm./sec.

0.55

Conductance, mmho/sq. cm. at 3 0 ° C .

79

Transport number, N a 2 5 ° C.

0.940

+

IN CHEMISTRY SERIES

in 0.37V N a C l at

New Anionic

0.95 176 0.034

As Discarded* Cationic Anionic 1

2.5 51 0.863

4.6 13 0.308

° Approximately 10-month equivalent continuous use.


Specifications required the following m i n i m u m properties (units as i n Table VII): Free electrolyte diffusion Conductance Transport number, t Mullen burst strength +

Not more than 1.3 Not less than 1.3 0.03 Not less than 25 p.s.i.

Conclusions Experience from 18 months' actual operation showed: A large-scale sheet flow type of electrodialysis plant can be operated satisfactorily and reliably when large membranes and large presses are used. E q u a l flow rates and balanced pressures i n diluting and concentrating compartments are necessary when the slotted gasket distribution system is used. If broad compartments are to be used, careful consideration must be given to the means of obtaining even flow distribution across a compartment and freedom from intercompartmental leakage. Gaskets should be soft enough to permit their thickness to be adjusted to that of the intermembrane spacer. In considering the size and number of units to be used, attention should be paid to the labor requirements for membrane replacement and pack maintenance. The below-capacity performance of the F S G plant was due to operation too near the critical polarization velocity, excessive intercompartmental leakage, and inability to obtain components of the required standard i n the limited time available for their development and manufacture. Acknowledgment T h e author is indebted to the South African Council for Scientific and Industrial Research, the Anglo American C o r p . of South Africa, L t d . , the Anglo Transvaal Consolidated Investment C o . , L t d . , R a n d Mines, L t d . , Johannesburg Consolidated Investment C o . , L t d . , N e w Consolidated G o l d Fields, and U n i o n C o r p . for permission to publish this paper. Thanks are due to J . R . W i l s o n , who was responsible for the design of the F S G electrodialysis plant and for the commissioning program. M u c h of the subject matter has been drawn from his thesis and other unpublished work. Thanks are also due to the manager of the Free State G e d u l d G o l d M i n i n g C o . , the staff of the Free State G e d u l d Water Demineralisation Plant, the staff of the Consulting C i v i l Engineering and Consulting Metallurgists Departments of the Anglo American Corp. of South A f r i c a , L t d . , and the members of the Process Development Division of the National Chemical Research Laboratory.


10. VOLCKMAN

Electrodialysis


Plant

157


Literature Cited (1) Moyers, W . H., Optima (Anglo American Corp., Johannesburg) (September 1957). (2) Moyers, W . H., Volckman, O . B., Trans. So. African Inst. Civil Engrs. 7, 309-24 (1957). (3) Office of Saline Water, "Standardized Procedure for Estimating Costs of Saline Water Conversion," March 1956. (4) Rapson, W . S., Optima (June 1955). (5) Volckman, O. B., Brit. Chem. Eng. 2, 146 (1957). (6) Volckman, O . B., "Electrodialysis Research and Development in South Africa," C C T A / C S A Specialist Conference on the Treatment of Water, Pretoria, Republic of South Africa, September 1960. (7) Volckman, O . B., Moyers, W . H., Natl. Acad. Sci.-Natl. Research Council, Publ. 568, 283-315 (1958). (8) Wilson, J. R., CSIR Contract, Rept. F . 31 (February 1960). (9) Wilson, J. R., ed., "Demineralization by Electrodialysis," Butterworths, London, 1960. (10) Wilson, J. R., "Depolarisation in Electrodialytic Demineralisation widi Particular Reference to the 100,000 gph Electrodialysis Plant Erected at the Free State Geduld Mine, Welkom, Republic of South Africa," Trans, Inst. Chem. Eng. (London) 41, 3 (February 1963). (11) Wilson, J. R., "Design and Commissioning of the World's First Commercial Scale Electrodialysis Plant," SACSIR Contract Rept. F . 31/1/60 (July 1960) and Supplement of March 1961, SACSIR Contract Rept. C F . 31/1961. R E C E I V E D May 28,

1962.


Saline Water ConversionâII

Recommend Documents

Saline Water ConversionâII